Assessing Italian News Reliability in the Health Domain through Text Analysis of Headlines

Fake news detection and fact checking represent challenging research areas in Natural Language Processing (NLP), especially in the health domain, which presents specific characteristics to be dealt with. On the one hand, online sources have become one of the main channels to retrieve health-related information. On the other hand, most of the time such online information suffers from lack of quality and requires domain-specific knowledge to be assessed. Therefore, the spread of untrustworthy health-related content urges to be mitigated since it may represent a threat for lives. To this aim, we develop a domain-specific annotated dataset suitable for training automatic systems to assess Italian news reliability. Our proposal tries to overcome some of the limitations of the available datasets by applying an in-depth text analysis to obtain a more fine-grained reliability assessment in the health domain.